svd-ai-lab/sim-cli/hypermesh

hypermesh-sim

Use this skill for visible, human-in-the-loop HyperMesh work through sim. The goal is a shared mesh-preparation loop, not unattended headless execution.

Preferred Loop

1. Check availability:

uv run sim check hypermesh

1. Connect with a visible desktop session:

uv run sim connect --solver hypermesh --ui-mode gui --workspace .\workspace
uv run sim inspect session.health

1. Execute one bounded step at a time:

uv run sim exec --file .\step.py --label mesh-step
uv run sim inspect last.result
uv run sim inspect hypermesh.workspace.files

1. Visually inspect the GUI after meaningful actions: import, cleanup, mesh

generation, quality checks, load/property assignment, and export.

1. Save or export checkpoints before moving to the next layer.

Snippet Rules

• Keep each snippet small enough to review from last.result.
• Write a JSON-serializable object to _result when possible.
• Prefer numeric acceptance data: node/element counts, quality criteria,

exported file paths, warnings, and failed entity counts.

• Do not rely on process exit alone as engineering acceptance.
• If the human changes the GUI, inspect health/results again before continuing.

Expected Result Shape

Useful snippets set _result, for example:

_result = {
    "ok": True,
    "stage": "mesh-quality",
    "nodes": 1200,
    "elements": 950,
    "max_aspect": 4.2,
    "exported": r"C:\work\case\mesh.fem",
}

The plugin wraps snippets and returns stdout, stderr, _result, elapsed time, and workspace artifacts through uv run sim inspect last.result.

Hard Stops

• If session.health.connected is false, reconnect before making changes.
• If a GUI dialog appears, resolve it visibly before continuing.
• If a mesh step produces no counts, no export, and no quality data, treat it

as incomplete even if the command returned successfully.